Long-Distance Paradox and the Hybrid Nature of Language Lorenzo G
Total Page:16
File Type:pdf, Size:1020Kb
23/8/2018 e.Proofing Long-Distance Paradox and the Hybrid Nature of Language Lorenzo G. Long-Distance Paradox and the Hybrid Nature of Language ✉ Guillermo Lorenzo, 1 Phone 0034985104653 provided by Repositorio Institucional de la Universidad de Oviedo View metadata, citation and similar papers at core.ac.uk CORE Email glorbroughte to you byn [email protected] 1 Department of Spanish Philology / Linguistics, Faculty of Humanities, University of Oviedo, C/ Amparo Pedregal s/n, E 33011 Oviedo, Spain Received: 7 February 2018 / Accepted: 18 August 2018 Abstract Nonadjacent or longdistance dependencies (LDDs) are routinely considered to be a distinctive trait of language, which purportedly locates it higher than other sequentially organized signal systems in terms of structural complexity. This paper argues that particular languages display specific resources (e.g. noninterpretive morphological agreement paradigms) that help the brain system responsible for dealing with LDDs to develop the capacity of acquiring and processing expressions with such a humantypical degree of computational complexity. Independently obtained naturalistic data is discussed and put to the service of the idea that the abovementioned resources exert their developmental role from the outside, but in compliance with other internal resources, ultimately compounding an integrated developmental system. Parallels with other human and nonhuman developmental phenomena are explored, which point to the conclusion that the developmental system of concern can be assimilated to cases currently been conceptualized as ‘cue response systems’ or ‘developmental hybrids’ within the ecological developmental paradigm in theoretical biology. Such a conclusion is used to support the idea that both current externalist and internalist concepts fall short of a correct characterization of language. Keywords Externalism http://eproofing.springer.com/journals_v2/printpage.php?token=FHrh5saHR9SPfdCyjnNGopism5m0Y0rDSKaA-qHVPMg 1/27 23/8/2018 e.Proofing Internalism Ecodevo biology Cueresponse mechanisms Developmental hybrids Introduction Longdistance—or nonadjacent—dependencies (LDDs) are pervasive in language, manifest in such diverse phenomena as discontinuous phrases across embedded constituents (1a), pronominal and anaphoric coreference (1b), wh word/gap or NP/gap connections in relative clauses (1c), interrogative (1d), and passive constructions (1e), etc. (Zushi 2013)1: (1) a. The man was saying [that his wife had been shot] with a desperate voice b. The woman1 that you2 visited is worried about herself1/your2 sons c. The man who1 she visited t1 yesterday is completely out of control d. Who1 did you visited t1 last night? e. She1 was declared guilty t1 In many accounts, LDDs are taken to be the hallmark of language uniqueness, as compared to the cognitive makeup of nonhuman species (Corballis 2011), or even in the context of human cognition (Hauser et al. 2002). The rationale for this tenet is that LDDs reveal that the powers the computational the powers of the computational resources of humans or language exceed anything hitherto known in other species or abilities. To begin with, the capacity for nested embedding— as in (1a), which illustrates the kind of recursion customarily associated to contextfreeness, is widely believed to be lacking in nonhuman cognition—see below. Among LDDs, special mention should be given to the family of cases routinely referred to as ‘agreement’ phenomena, in which an item adopts a morphological appearance that mirrors the feature composition of another designated one. Agreement covers, from an interlinguistic point of view, an extremely wide range of related phenomena, which may be epitomized by the relatively simple case of subject agreement in English. In this language, as in many others, it is common for the verb and the agreeing subject to appear as adjacent units in the linear http://eproofing.springer.com/journals_v2/printpage.php?token=FHrh5saHR9SPfdCyjnNGopism5m0Y0rDSKaA-qHVPMg 2/27 23/8/2018 e.Proofing order of utterances—as in (2a); however, (2b) clearly illustrates the LDD character of the relation: (2) a. The girl1 wants1 to know who is afraid of her b. The girl1 who is afraid of the boy wants1 to know who is afraid of her Crucially, note that in (2b) there is a matching item that is closer (actually, adjacent) to the relevant verb—namely, the boy, which however does not qualify as a better candidate for agreement than the sixitemdistant correct one—the girl. Note, also, that (2b) further illustrates that different agreement relations may concur and interfere in a single utterance—namely, the girl is also the semantic controller of the abstract subject of to know, as well as the antecedent of the pronoun her. In each of these different instances of LDD, the second relatum is increasingly more distant from the first one, without hindering the interpretation of the whole utterance. These kinds of observations have served to rank natural language in a relatively high position within the standard Chomsky hierarchy of computational complexity (Chomsky 1956): Namely, an area above the contextfree recursive languages—i.e. those that accept discontinuous dependencies created by embedding segments within larger segments, since natural language accepts dependencies that hold across segments belonging to different levels of embedding. (3a) offers a suitable illustration, which is shown as an abstract pattern in (3b)—different capital letters represent that units belong to different embedded segments: (3) a. The girl1/3 who said that the boy2 hated her3 loves1 him2 b. X1/3 … Y … Z2 … Z3 … X1 X2 The reason why contextfreeness is customarily seen as critical in the literature on comparative cognition is the more or less consensual claim that no clear cases of bona fide structural embedding have to date been found in any example of nonhuman sequential behavior whatsoever, while human language appears to be situated even some steps ahead of such a level of computational complexity. Context freeness thus represents a kind of chasm between human and nonhuman cognition (see Fitch and Hauser 2004; van Heijningen et al. 2009; Beckers et al. 2012). http://eproofing.springer.com/journals_v2/printpage.php?token=FHrh5saHR9SPfdCyjnNGopism5m0Y0rDSKaA-qHVPMg 3/27 23/8/2018 e.Proofing Agreement dependencies are worth special mention in this context because, unlike other kinds of LDDs, they exhibit the peculiar trait of not being interpretable, in the sense that they make no specific contribution to the compositional meaning of the sentence (Chomsky 1995). Rather, agreement merely involves associating some morphological material to an item, which mirrors the (interpretable) feature composition of another (potentially distant) item—e.g. –s in not required to contribute the meaning ‘subject, 3rd person, singular’ in (2a), which is already contributed by the constituent the girl itself. For this reason, it is also commonly said that agreement is an asymmetric relation, since it is the verb that agrees with the subject, but not the other way around; in other words, the subject is a meaningful constituent, but the subject agreement bearer item affixed to the verb is meaningless. The fact that agreement features are in many cases uninterpretable replicas of interpretable ones is certainly confounding (Adger and Svenonius 2011). Some features function both within interpretable and uninterpretable bundles in many languages; moreover, homologues of certain features may be interpretable in some languages and uninterpretable in some others (Van Valin 2003). In any event, the kind of features and morphology that I am interested in here are those displayed in LDDs without any conceptual import. Such property raises an important question, to which no clear, consensual answer exists to date: Why are agreement relations so pervasive, if they do not appear to make any clear meaningful contribution to utterances? This is one of the main issues on which this paper will focus. I shall specifically defend that neither cognitive/functionalist approaches (Barlow and Ferguson 1988) or formalist/generativist counterparts (Boeckx 2006) have hitherto provided an account of the ultimate raison d’être of agreement systems in natural language. In a nutshell, functionalist/cognitive approaches try to make sense of agreement by ascribing some communicative role to its bearers (e.g. fixing discourse referents, creating noisepreventing redundancies, etc.), while formal/generativist approaches conceptualize it as a kind of imperfection that the language system needs to get rid of when processing expressions. The alternative that I shall suggest is a developmental one, which is not incompatible with prior suggestions along the lines of those referred to above, but which is more basic and closer to a ‘first cause’ for the existence of such linguistic entities. Another point of interest of this paper has to do with a further characteristic of agreement relations, which illustrates an aspect of language in a way that other kinds of LDDs do not, or, at least, not as straightforwardly. The ways in which agreement relations instantiate in different languages are manifold and sensitive to a very wide range of criteria. To cite a few, agreement may be sensitive to a http://eproofing.springer.com/journals_v2/printpage.php?token=FHrh5saHR9SPfdCyjnNGopism5m0Y0rDSKaA-qHVPMg 4/27 23/8/2018 e.Proofing particular syntactic relation—e.g. subject agreement, as in (2) above; or sensitive to a particular morphological feature—e.g. number agreement, as in (4), where the affix may indistinctly refer to the subject or the object, which must be a plural one; or sensitive to complex combinations thereof—as in (5), where the affix responds to a person/number hierarchy (1st/2nd > 3rd pl. > 3rd sg.), but only when the agent is focused and the transitive verb turned into an intransitive: (4) Georgian (from Nevins 2011) gxedavt. 2objsawpl ‘I saw you all; we saw you all; he saw you all; we saw you’ (indistinctly) (5) Quichean (from Preminger 2014) a. ja yïn xinaxan ri achinfoc me asp1sg.abshearaf the man ‘It was me that heard the man’ b.